Using Metadata to Maximize Yield and Expand Inventory in TV - Contextual Advertising
PFT Blog Team | 08 Sep 2016

Using metadata to maximize yield and expand inventory in TV by leveraging additional metadata layer of in-video context in a way that enables broadcasters to optimally sell their video inventory and brands to place their ad in the right program at the right time and place
Using metadata to maximize yield and expand inventory in TV by leveraging additional metadata… Click To Tweet

Today, consumption of video over the internet is the new normal. But while online video viewing is getting increasingly personalized, the ads that play in and around the video are still far from being personal and relevant. As brands increasingly become conscious of the ROI on every ad dollar spent, broadcasters need to adopt new ways to ensure the effectiveness of each ad played out on their network.

Ramki Sankaranarayanan, PFT’s Founder & CEO, presented an innovative solution to address this challengeat IBC 2016 - one of the world’s largest gatherings of technology professionals from the Media & Entertainment industry.The focus of his presentation was a completely new approach to Contextual Advertising, which involves leveraging an additional metadata layer of in-video context in a way that enables broadcasters to optimally sell their video inventory and allows brands to place their ad in the right program at the right time and place.

Typically, Contextual Advertising is a function of global attributes only. From the content perspective, parameters like the content type, title, genre, language etc. are usually considered. While from a user profiling standpoint, parameters like age, gender, and location dominate. The traditional method of marrying these two sets of parameters to decide which ad to show, either pre-defined or real time, is limiting for both the publisher and the advertiser. For the publisher, the volume of inventory is limited and yield per unit of inventory is sub-optimum. For the advertiser, he/she might be missing out on many relevant points during the video where serving the same ad would make more sense. Additionally with limited “context”, the ROI on ad spends remains ineffectual, as there is no way to know if the right ad is being shown to the right target group at the right time! Needless to say, there is a pressing need for a solution that increases the ad inventory as well as its relevance without impacting the viewing experience.


The new approach to Contextual Advertising involves using in-video context like the mood, emotion or action of a scene/character to create relevant keywords or metadata, either for an exact point or a given time interval within a video. When coupled with already available content (e.g. title, genre etc.) and user (e.g. age, location etc.) metadata, it becomes highly enriched information, which can then be passed on to the ad decision system. Owing to the additional layer of in-video metadata that has come in, the ad system is able to make a better decision of which ad to show. This in turn increases its relevance for that in-video ‘moment’, and results in better ROI for advertisers. Also, since each frame/time interval is described using certain metadata, the number of such ‘moments’ where an ad can be shown goes up, providing more opportunities for a publisher/broadcaster to show an ad.

This approach involves the addition of just two steps to the content and ad-serving ecosystem. The first step is to create a data model which defines the in-video information that needs to be captured. This involves the extraction of metadata using a combination of automated tools and manual effort. Automated tools analyze key frames within the video, taking into account variables like facial expressions, objects, background, foreground and color depth. The extracted information is then compared against an existing database of information (images, text etc.) for a match, resulting in a set of descriptors/keywords for a given in-video ‘moment’. As automated tools may generate a lot of metadata which may be irrelevant for a particular context, a layer of manual metadata curation and tagging needs to be added on to get the final, cleaned up and pertinent metadata. Next, the extracted metadata is passed onto an interpretation engine, which processes the rich metadata to generate markers that show potential ‘moments’for placing an ad. Each such‘moment’ is mapped to a tag within the ad system for custom targeting. For instance, if the metadata extractor has described a scene as a “car accident”, the interpretation engine would map that information to an ad tag like “insurance” within the ad system.

This ground-breaking solution creates a win-win situation for both publishers as well as advertisers. For publishers, using in-video context metadata increases the opportunities to show an ad, as they can conveniently slice and dice content based on in-video context, and create packages that can be sold in a targeted manner to multiple advertisers. Further, since the inventory being made available is richer in context, it can also command higher ad rates. For advertisers, showing an ad at the most opportune moment increases its relevance tremendously.This results in improved brand affinity and better ROI. Most importantly though, the end consumer gets a seamless experience of content and ads bundled together that is non-intrusive and does not disrupt the overall viewing experience.

Click here and we'll get you a meeting with our Subject Matter Expert


Subscribe to get emails delivered to your inbox

Email *

Thank you!
Our team will get in touch with you shortly.